IBM has begun offering its coding-agent platform "Bob" as a self-hosted deployment that runs in on-premises, sovereign, and air-gapped environments. We break down this move — along with its prerequisites and open questions — that hands regulated industries unable to let code leave the building the o
IBMがコーディングエージェント基盤「Bob」をオンプレ・ソブリン・エアギャップ環境で動かせるセルフホスト型として提供開始。コードを社外へ出せない規制産業に、AI開発を自社の壁の内側で回す選択肢を差し出す一手を、前提と論点まで整理します。
Google has added "Bring Your Own Agent" to Android Studio, letting you plug in Claude, Codex, and Antigravity via the shared ACP standard and run them on your own subscription. Here are the key points—and the cautions—of an era where you keep the IDE but swap only the agent.
GoogleがAndroid Studioに「Bring Your Own Agent」を追加。Claude・Codex・Antigravityを共通規格ACP経由で差し込み、自分の契約のまま走らせられる。IDEを固定したままエージェントだけ乗り換える時代の勘所と警鐘を整理する。
Coding agents dumped "review screenshots" onto public GitHub, leaking over 13,000 internal images from more than 300 companies. A look at "PixelLeak" — where the hole wasn't an attack but an agent's own self-made workaround — and the brakes you can apply right now.
コーディングエージェントが「レビュー用スクショ」を公開GitHubに放流し、300社超・1万3千枚超の内部画像が流出。攻撃ではなくエージェントの自発的な抜け道が穴になった「PixelLeak」と、いま踏めるブレーキを整理する。
Claude Code v2.1.287 introduces "Mods," running JS/TS handlers inside the core process so they can replace everything from tool calls to screen rendering. They're on by default, but they run with the same privileges as Claude Code itself — making it essential to vet your distribution sources.
Claude Code v2.1.287 が「Mods」を導入。JS/TSのハンドラを本体内部で走らせ、ツール呼び出しや画面描画まで差し替え可能に。既定で有効だが権限はClaude Code本体と同じで、信頼できる配布元の見極めが要る。
Codex CLI v0.160.0 was released on October 1, 2026. The approver Guardian can now read agent-to-agent handoffs and the original instruction, and new additions include a browsable command seat and the ability to start outside a project. Here's a rundown of the improvements that help unattended, multi
Codex CLI v0.160.0が2026年10月1日公開。承認役Guardianがエージェント間の引き継ぎと最初の指示を読めるようになり、履歴をたどれる指令席やプロジェクト外からの起動も追加。無人・多段運用に効く改良と、その注意点を整理します。
On September 28, Anthropic added build-eval and hillclimb to Claude Code's claude-api skill. They semi-automate eval creation and self-improvement, with built-in safeguards against overfitting, replacing "tuning by instinct" with tuning by the numbers.
Anthropicが9月28日、Claude Codeのclaude-apiスキルにbuild-evalとhillclimbを追加。評価づくりと自己改良を半自動化し、過学習を自分で疑う仕組みで「勘のチューニング」を数値に置き換える。
